| from transformers import PretrainedConfig | |
| class SpinConfig(PretrainedConfig): | |
| model_type = "spin" | |
| attribute_map = { | |
| "hidden_size": "d_model", | |
| "num_hidden_layers": "n_layers", | |
| "num_attention_heads": "n_heads", | |
| } | |
| def __init__( | |
| self, | |
| vocab_size: int = 512, | |
| max_seq_len: int = 256, | |
| d_model: int = 48, | |
| n_layers: int = 2, | |
| n_heads: int = 4, | |
| d_ff: int = 128, | |
| norm_eps: float = 1e-5, | |
| tie_word_embeddings: bool = True, | |
| **kwargs, | |
| ): | |
| self.vocab_size = vocab_size | |
| self.max_seq_len = max_seq_len | |
| self.d_model = d_model | |
| self.n_layers = n_layers | |
| self.n_heads = n_heads | |
| self.d_ff = d_ff | |
| self.norm_eps = norm_eps | |
| self.hidden_size = d_model | |
| self.num_hidden_layers = n_layers | |
| self.num_attention_heads = n_heads | |
| super().__init__(tie_word_embeddings=tie_word_embeddings, **kwargs) |